How Facial Recognition Software Works
Facial recognition is something that comes easily to humans – we learn it within weeks of birth – but it was a much harder feat to achieve with a computer. We recognize faces because we are, essentially, ‘programmed’ to do so by our evolutionary psychology.
However, a computer – which is based on a structure of logical, mathematical principles -doesn’t have the emotional input that allows us to recognize even a colon and an end-bracket as a smiley face.
To get a computer to recognize a face you have to program it to look for certain things. For a start it has to be able to distinguish between a human face and everything else. This is done by programming a piece of software to look for what are known as ‘nodal points’. These statistics govern things like the common width of noses or the standard range of depth for eyes.
Once the software can recognize the basic concept of a face, it can be programmed to recognize certain people. Your camera may have a Face Recognition feature, which you have to initially set up, taking several pictures of people and tagging them by name. The software learns the measurements of the nodal points around their faces thus recognizing them automatically.
1. Finding the face – A security camera captures a shot of someone’s face in a crowd, which is fed back to a central security computer in the CCTV network.
2. Image breakdown – The computer breaks the image down into components, then into data, which it reads.
3. Recognize features – The computer recognizes patterns and features it has been programmed to look for.